Re: [Discuss-gnuradio] File sink: file size

[Marcus D. Leech]

On 04/07/2017 02:53 PM, Ellie White wrote:

Hello,

As I've mentioned in an earlier email to this list, I'm working on a small loop antenna with the purpose of monitoring solar activity. I am currently reading in the data through an RTL-SDR source, flowgraph attached.

For this project, I need to be able to let my antenna run for long periods of time (upwards of an hour or more, preferably several hours) in order to get a graph of signal intensity vs. time. Right now, my setup consists of the GNU Radio flowgraph attached, that produces a file which I analyze using a Python program to get a spectral plot and a time series (for one of the 1024 channels saved in the file).

I let my antenna and GNU Radio run for about 15 minutes last night collecting data, and the file size ended up at over 7 GB. I am wondering if there is any way to reduce the size of the file without tremendously reducing the quality of the data? I have a few ideas on this, but I'm hoping someone else will have better ones:

-Maybe I need to reduce the sample rate.

-Another option is to somehow just save the channel I need to the file, instead of all 1024. Does anyone have any advice as to how to accomplish that?

If you need any more information or if I should clarify anything, please let me know. Thanks!

Ellie

_______________________________________________
Discuss-gnuradio mailing list
address@hidden
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio

You might also take a look at my work from 6 or 7 years ago:

https://github.com/patchvonbraun/SIDSuite

It uses audio input.  It uses a set of Gortzel transforms to pick out the channels of interest.

But there are other approaches, using an FFT as you have. 

Since the things that you're looking for unfold over timescales of many-seconds to many-minutes, you don't need to be logging your FFT output in
  real-time.  You could easily integrate for several seconds and dump output data at a lower rate, using keep-one-in-N.

At an input rate of 2Msps, your 1024-point FFT will be producing 1953 1024-bin vectors/second.  This is *much* faster than the timescale of ionospheric
  events, in my experience.  Also, your frequency resolution will be somewhat coarse-1953Hz.  This is much wider bandwidth than the naval stations
  you're looking at, so you'll mostly be integrating noise, rather than navy signal, in your bins.